06/10/2026
Language models at work, carburetors in the garage. For Johan Östling, our Head of Data and AI, both come down to the same thing: understanding how something works before you try to do anything with it.
Looking under the hood is also what changed his mind about AI. For years, he expected bigger and better models to eventually handle everything on their own. He doesn't think that anymore. What decides how good the output is comes down to everything around the model: the data you give it, how you write your prompt and the context you add.
In practice, two people using the same model can get very different results. Usually, the one who gets the better answer is the one who explained what they actually needed in a clear way.
We sat down with Johan for the first interview in our new Q&A series. He talks about where the vibe coding debate goes wrong, the first thing he checks in any new AI tool and the one piece of advice he'd written down before the interview even started.
Read the full conversation here: https://eu1.hubs.ly/H0ySMvp0